The scale of the rescue effort is remarkable. During two field campaigns in 2023, a team led by Derrick Muheki of the Vrije Universiteit Brussel first spent two weeks compiling a comprehensive inventory of the archives, cataloguing station names, codes, coordinates and altitudes, and verifying that each bundle of sheets was correctly assigned to its station, year and month. They then spent two months photographing every available sheet using a digital camera mounted on a tripod inside a portable photo box with uniform lighting. Because the archive site has no electricity, the entire setup ran on rechargeable batteries. In total, 9885 sheets from 37 weather stations across the Democratic Republic of the Congo were imaged, covering records that mostly span from the 1960s to the mid-1990s, with a few extending into the early 2000s. The urgency was real: a fire had damaged one wing of the Yangambi archive just months before the digitization campaign began.

Photographing the sheets was only the first step. Reading them posed a far greater challenge, since every value was handwritten, in styles that varied from observer to observer and decade to decade, on paper of wildly varying quality and condition. To solve this, the team applied an improved version of MeteoSaver, an open-source software package they developed that uses machine learning to transcribe handwritten tabular data. The software processes each image through a pipeline of modules: it enhances the image with adaptive threshold binarization, detects the table structure and individual cells using computer vision algorithms, recognizes the handwritten values with a trained Tesseract optical character recognition model, and then subjects every transcribed value to a battery of automated quality control checks before formatting the results in the standardized Station Exchange Format prescribed by the Copernicus Data Rescue Service.

The quality control system is where the technical ingenuity really shows. Each transcribed value is checked against region-specific temperature thresholds, logical consistency rules and inter-variable relationships. Because the original observers calculated five-day and monthly totals and averages on every sheet, the software can recalculate these summaries from the transcribed daily values and compare them with the handwritten ones, confirming entire blocks of data when they agree within a 0.2 degree Celsius margin that accounts for historical rounding. The sheets also record dry bulb temperature, wet bulb temperature, actual vapour pressure, relative humidity and vapour pressure deficit at 06:00, 15:00 and 18:00 local time, and these variables are physically linked by well-known atmospheric equations. The software exploits those relationships, using a modified Magnus-Tetens formula and the psychrometric equation from the Smithsonian meteorological tables, to cross-validate the transcribed values and flag or correct inconsistencies. This iterative approach paid off dramatically: after five consecutive runs over the archive, in which newly confirmed transcriptions were fed back into the training data, the number of quality-controlled confirmed values rose from about 321,000 to more than 1,019,000, a 215 percent increase.

Validation against independently hand-transcribed records from five stations showed that the automated pipeline achieved match rates between 81.2 and 92.3 percent, with a median of 87.4 percent, and a median mean absolute error of 0.35 degrees Celsius. That error is of the same order as the observational uncertainty of historical thermometer measurements in the tropics, and part of the disagreement likely stems from errors in the manual reference data itself rather than the machine. Most mismatches were not misread digits but one-day shifts caused by row misalignment in table detection, a problem the authors identify as a priority for future improvement. Combined with previously transcribed datasets, including the COBECORE project’s pre-1960 records from the Belgian State Archives and a manually transcribed series for Yangambi-Km.5 extending to 2020, the new dataset spans 1908 to 2023 and forms the largest in situ temperature and precipitation dataset ever assembled for the Congo Basin.

What the rescued records reveal is unambiguous. Of the 21 stations with sufficient data for trend analysis, 20 show warming in at least one of daily maximum, minimum or average temperature. The median trends are 0.22, 0.10 and 0.15 degrees Celsius per decade respectively, corresponding to roughly 0.7, 0.3 and 0.5 degrees Celsius of warming over a 30-year period. These magnitudes are consistent with previous regional assessments that estimated 0.75 to 1.05 degrees Celsius of warming in Central Africa since 1960, but they now rest on direct, station-level evidence from within the basin itself. The analysis, based on the Theil-Sen estimator, a robust method for detecting trends in noisy time series, also shows that the temperature distributions themselves are shifting: with each successive decade between 1961 and 1990, the probability distributions of daily maximum, minimum and average temperatures move progressively toward warmer values, with modal temperatures and 95th percentile extremes rising by nearly a degree Celsius between the first and last decades at several stations.

The changing face of temperature extremes is perhaps the most striking finding. Across the aggregated station network, the average number of hot days per year, defined as days exceeding the 95th percentile of daily maximum temperature, nearly doubled between the earliest and most recent decades of the common observation window, rising from about three days per year above 33.3 degrees Celsius to about five. Days with average temperatures above 26.9 degrees Celsius doubled from roughly three to six per year. Meanwhile, cold days, those falling below the 5th percentile, declined from around seven to five per year, with cold nights showing a similar decline. At individual stations with longer records, such as Luki-Plateau, Mulungu-Mulohe and Yangambi-Km.5, the number of hot days more than tripled over the period 1961 to 2000. Because the IPCC’s Sixth Assessment Report could not assess changes in hot extremes in Central Africa at all, these observations fill a genuine hole in the global picture of climate change, providing the raw material that future detection and attribution studies will need.

The precipitation story is more tentative. Only three stations had more than 30 years of quality-controlled daily rainfall data, and at two of them, Yangambi-Km.5 and Mulungu-Mulohe, heavy precipitation events became markedly more frequent in the later half of the record. At Mulungu-Mulohe, a daily rainfall of 63 millimetres, which had a return period of about two years in the earlier period, became roughly a once-a-year event in the later period. At Yangambi-Km.5, an 83-millimetre day likewise became twice as frequent. The third station, Gimbi-Plateau, showed no substantial change. With such limited spatial coverage, the authors are careful to present these as station-specific examples rather than a representative picture of rainfall change across the Democratic Republic of the Congo, a caution that reflects the harder quality control problem posed by precipitation, for which fewer automated consistency checks are available.

The team also compared their station trends with the ERA5-Land reanalysis, a state-of-the-art gridded product from the European Centre for Medium-Range Weather Forecasts. The reanalysis captures the seasonal cycle and the broad warming trend from the 1960s onward reasonably well, but shows no systematic pattern of agreement or disagreement with the stations, sometimes underestimating and sometimes overestimating local trends. More troubling, ERA5-Land exhibits a pronounced artificial jump of about 4.5 to 5 degrees Celsius in its temperature series during the early 1950s that appears in none of the station records from that era, a sign that the reanalysis is weakly constrained in this region during its early period when few observations were available to anchor it. The rescued dataset, now uploaded to open-access repositories including the Copernicus Global Land and Marine Observations Database and Zenodo, offers exactly the kind of ground truth needed to fix such products.

The implications reach well beyond Central Africa. Most station records end in the mid-1990s, when the First and Second Congo Wars disrupted observations nationwide, meaning the full acceleration of warming since the 1990s is likely not captured in these data. Roughly 300,000 additional temperature records and 200,000 precipitation records were transcribed but did not yet pass quality control, and 225 hand-drawn sheets were excluded entirely, all of which could be recovered with improved software or citizen science verification. Meanwhile, the COBECORE project has imaged records from 575 pre-1960 stations that could add millions more observations if transcribed. The authors present their workflow, from inventory to imaging to machine-learning transcription to quality control, as a replicable template for other data-sparse regions where millions of weather observations still moulder on paper. For a basin that holds the world’s second-largest river and rainforest, and some of its most climate-vulnerable populations, turning those fragile pages into hard numbers is not just an archival triumph; it is the foundation for understanding, and preparing for, the region’s warming future.

Subject of Research: Climate data rescue and observed warming trends in the Congo Basin

Article Title: From paper to proof: revealing Congo Basin warming through rescued climate archives

Article References: Muheki, D., Hufkens, K., Jacobsen, K., Verbeeck, H., Boeckx, P., Kankonde Ntumba, D., Kapalay Moulasa, O., Vercruysse, B., Birkholz, J. M., Verbruggen, C., Hawkins, E., Lampe, S., Yakusu, E. K., Imwangana, F. M., Mbifo, J., Likwela, T. B., Meunier, F., Dewitte, O., Thorne, P., & Thiery, W. (2026). From paper to proof: revealing Congo Basin warming through rescued climate archives. Earth System Dynamics, 17(5), 1455-1511. https://doi.org/10.5194/esd-17-1455-2026

Image Credits: AI Generated

DOI: 10.5194/esd-17-1455-2026

Keywords: Congo Basin, climate data rescue, Democratic Republic of the Congo, temperature trends, extreme heat, machine learning, MeteoSaver, historical weather records, ERA5-Land, precipitation extremes, IPCC, Earth System Dynamics

The research team, led by Kutubuddin Ansari and Emine Tanır Kayıkçı of the Department of Geomatics Engineering at Karadeniz Technical University in Trabzon, approached the problem with a deliberately standardized methodology. Rather than comparing raw temperatures or rainfall totals, which vary enormously between a coastal station on the Aegean and a high plateau site in eastern Anatolia, the authors converted every measurement into a standardized anomaly using Z-transformation. This statistical technique expresses each daily or annual value as a departure from the long-term mean, scaled by the standard deviation of the record at that station. The result is a dimensionless quantity that allows a fair comparison of climate behavior across radically different geographic settings, from the humid Black Sea coast to the semi-arid interior of Central Anatolia.

With the anomalies in hand, the researchers computed long-term trends for four surface variables: temperature, sea-level pressure, wind speed and precipitation. The signal that emerges most clearly from the data is a nationwide warming tendency, one that is particularly pronounced in central and western Türkiye. This spatial pattern is consistent with the country’s position at the intersection of several vulnerable climate zones. Türkiye sits at the eastern edge of the Mediterranean basin, a region that climate model projections have long identified as a hotspot where warming is expected to outpace the global average, and it simultaneously experiences continental influences from the Anatolian interior and maritime influences from the Black Sea, the Aegean and the Mediterranean. The new observational analysis confirms that the warming signal is not confined to one corner of the country but is a genuinely national phenomenon, with its strongest expression in the interior and western provinces.

The second robust signal in the record is a decline in wind speed, a finding that echoes what atmospheric scientists have called global terrestrial stilling. Studies published over the past fifteen years have documented widespread decreases in near-surface wind speeds across many continents, with proposed explanations ranging from increases in surface roughness driven by vegetation growth and land-use change to shifts in large-scale atmospheric circulation. The Turkish data add another national record to that growing body of evidence. For a country that has invested heavily in wind power capacity along its Aegean and Marmara coasts, the long-term behavior of surface winds is not an academic curiosity; it bears directly on the reliability of one of the fastest-growing segments of the national energy system. The study’s finding of an overall decline in wind speed across the 1946 to 2025 period will therefore be scrutinized closely by both climatologists and energy planners.

Not every variable in the analysis tells such a decisive story. Sea-level pressure and precipitation both exhibit trends that are weaker and far more regionally variable than those for temperature and wind. This is in many ways exactly what one would expect. Precipitation over Türkiye is governed by a complicated interplay of Mediterranean storm tracks, Black Sea moisture fluxes, orographic lifting over the North Anatolian mountains and the Taurus range, and the seasonal migration of the subtropical high. Averaging such a heterogeneous set of processes into a single national trend risks obscuring the local signals, and the study’s authors appear to have taken that heterogeneity seriously by reporting the regional variability rather than forcing a single narrative onto the rainfall data.

One of the study’s most technically interesting results concerns the relationships among the surface variables themselves. When the researchers computed correlations between temperature, pressure, wind and precipitation anomalies, they found that the variables are generally only weakly correlated with one another. This weak internal coupling is a meaningful diagnostic. It indicates that each variable responds to a different combination of regional atmospheric circulation patterns and local geographical controls, such as elevation, distance from the sea and topographic exposure. In a strongly coupled system, a change in circulation would move temperature, pressure, wind and rainfall together in predictable ways; in the Turkish record, each variable appears to march to its own drummer to a considerable degree, which complicates any attempt to reconstruct one variable from another or to attribute all observed changes to a single driver.

The heart of the paper, however, lies in its search for teleconnections, the statistical fingerprints that distant oscillations leave on regional climate. The authors examined three candidates. The first, the Quasi-Biennial Oscillation, is a stratospheric phenomenon in which equatorial winds alternate between easterly and westerly phases on a cycle of roughly 28 months, descending through the stratosphere from about 10 hPa down to about 70 hPa and beyond. The QBO is known to modulate the polar vortex, the Madden-Julian Oscillation and, in some studies, monsoon systems, so a link to mid-latitude surface climate is physically plausible. Yet when Ansari and Kayıkçı tested the QBO at both the 10 hPa and 70 hPa levels against their Turkish surface records, they found no statistically significant relationship with any of the four climate variables. For Türkiye, at least, the stratospheric biennial pulse appears to be a non-player at the surface.

The story changes when the analysis turns to the North Atlantic Oscillation and the Arctic Oscillation. The NAO describes a seesaw in atmospheric mass between the Azores high and the Icelandic low, while the AO captures the annular mode of the Northern Hemisphere, a ring of alternating pressure between the Arctic and the mid-latitudes. Both are dominant modes of extratropical variability, and both steer storm tracks and air masses across Europe and the eastern Mediterranean. The annual correlation analysis revealed a significant positive relationship between the NAO and precipitation, and between the AO and sea-level pressure, over Türkiye. In other words, when the NAO is in a positive phase, Turkish precipitation tends to be higher in the annual mean, and the AO leaves a clear positive signature in the country’s surface pressure record. These are exactly the kinds of relationships one would anticipate from the physics of mid-latitude circulation, in which positive NAO phases strengthen westerly flow and shift storm activity in ways that alter moisture delivery to Anatolia.

To quantify how much of the Turkish climate variability these circulation modes can actually explain, the researchers built combined regression models using the NAO and AO indices together as predictors. For sea-level pressure, the combined NAO-AO model explained 8.0 percent of the variance, a result the authors report as statistically significant. Eight percent may sound modest, and in absolute terms it is, but for annual mean surface pressure over a region as topographically and climatologically complex as Türkiye, a statistically robust eight percent is a genuine signal. It confirms that the mid-latitude annular modes are relevant players in Turkish climate variability, even if they are far from the whole story. By contrast, the corresponding regression models for temperature, wind speed and precipitation were weak or only marginally significant, underscoring that local and regional factors, along with the broader warming trend, dominate those variables more than any single teleconnection index can.

The broader significance of the study lies in its careful separation of what matters and what does not for Turkish climate variability. Over an 80-year window, the dominant signals are a clear warming trend concentrated in the central and western regions, a widespread stilling of near-surface winds, and a statistically detectable but partial influence of the NAO and AO on pressure and rainfall. The QBO, despite its importance in stratospheric dynamics and its documented teleconnections elsewhere in the world, drops out entirely at the surface over Türkiye. For climate modelers working on the eastern Mediterranean, these results provide a benchmark for evaluating which modes of variability regional simulations must capture, and for water managers, energy planners and agricultural stakeholders in Türkiye, the message is that the country’s climate future will be shaped less by exotic stratospheric oscillations than by the relentless background warming and the shifting behavior of the Atlantic-centered circulation patterns on its doorstep.

Subject of Research: Long-term climate variability and teleconnection patterns over Türkiye

Article Title: Long-term climate anomalies in Türkiye and their relationships with the QBO, NAO and AO

Article References: Ansari, K., & Kayıkçı, E. T. (2026). Long-term climate anomalies in Türkiye and their relationships with the QBO, NAO and AO. Theoretical and Applied Climatology, 157(10), Article 613. https://doi.org/10.1007/s00704-026-06541-z

Image Credits: AI Generated

DOI: 10.1007/s00704-026-06541-z

Keywords: Türkiye, climate change, temperature trends, wind speed, precipitation, North Atlantic Oscillation, Arctic Oscillation, Quasi-Biennial Oscillation, teleconnections, sea-level pressure, GSOD stations, Theoretical and Applied Climatology

The team’s starting point was deceptively simple: measure how variable the climate actually is, and whether its long-term trends are statistically real. To do this, they deployed a suite of well-established indices. The Coefficient of Variation captured how much rainfall swings from year to year relative to its average, while the Precipitation Concentration Index revealed whether rain arrives in destructive bursts or spreads evenly across the season. The Standardized Anomaly Index placed each individual year in context, flagging those that fell far above or below the long-term norm. For trend detection, they used the Mann–Kendall test, a non-parametric method that identifies monotonic trends without assuming the data follow a normal distribution, paired with Sen’s slope estimator, which quantifies the magnitude of change per year while remaining robust to outliers.

The rainfall findings are sobering. Precipitation in the study area proved highly variable both in space and in time, and in 58 percent of the years examined, rainfall fell below the long-term mean. That means for more than half of the past three decades, farmers in the central highlands have been farming under rainfall deficits, planning their planting and fertilizing around water that never arrived in expected quantities. High values of the Precipitation Concentration Index in certain years pointed to rainfall arriving in concentrated, intense episodes rather than gentle, sustained showers, a pattern that simultaneously raises erosion risk and reduces the fraction of water that actually infiltrates the soil profile where wheat roots can reach it.

Temperature told an unambiguous story. Both daily maximum and daily minimum temperatures rose significantly across all study locations between 1989 and 2020, with the researchers reporting statistical significance at the p < 0.01 level. Maximum temperatures climbed by 0.043 to 0.081 degrees Celsius per year, while minimum temperatures increased by 0.034 to 0.094 degrees Celsius per year. Compounded over the 31-year record, those rates translate into warming on the order of 1.3 to 2.5 degrees Celsius, a substantial shift for a highland ecosystem where wheat is already grown near the upper edge of its thermal tolerance in many locations. Rising minimum temperatures are particularly consequential for wheat, because warmer nights accelerate phenological development, shorten the grain-filling period, and increase respiratory losses that erode final yield.

Perhaps the most striking result concerns the growing season itself. Using INSTAT+ 3.37 statistical software, a tool specifically designed for agrometeorological analysis, the researchers characterized when the rainy season actually begins and ends, and how long the window for crop production lasts. They found that the start of the rainy season now ranges from June 13, corresponding to day 165 of the year, to as late as July 6, day 188. The end of the season ranges from October 10, day 284, to November 2, day 307. The length of the growing period consequently spans a wide band, from just 96 days to as many as 136 days. That 40-day spread is not an academic curiosity; it is the difference between a crop that completes grain fill before the rains vanish and one that fails mid-season.

This shifting crop calendar has direct, cascading consequences for bread wheat, a staple that supplies a large share of calories for Ethiopian households and dominates the highland farming systems of the study region. When the rains arrive late, farmers who plant on traditional dates sow into drying soil, and germination suffers. When the season ends early, wheat planted even a few weeks behind schedule encounters terminal moisture stress during flowering and grain filling, the growth stages most sensitive to water deficit. The study’s authors note that these climate shifts disrupt farming calendars, trigger crop failures, and threaten food security, a chain of impacts that begins with a few weeks of rainfall timing and ends on the household plate.

To quantify how much of the yield story climate can explain, the team turned to combined multiple regression analysis, modeling wheat yield as a function of the assessed climate variables across the selected study sites. The results revealed a striking spatial gradient in climate sensitivity: the climate variables explained between 17 and 83 percent of the change in wheat yield depending on the site. At the high end of that range, more than four-fifths of year-to-year yield variation is attributable to climatic factors alone, before accounting for soil quality, fertilizer use, variety choice, or pest pressure. The wide range itself is informative, suggesting that some locations buffer climate stress better than others, whether through deeper soils, higher elevations, or more favorable rainfall regimes, while others sit squarely on the climate knife-edge.

The implications extend well beyond the central highlands. Ethiopia is the largest wheat producer in sub-Saharan Africa, and wheat is a national priority crop in ambitious government self-sufficiency programs. Yet the study’s findings suggest that the biophysical foundation of that production is moving. A growing season that once reliably spanned four months can now contract to barely three, and nighttime temperatures that once supported slow, steady grain development are rising faster in some locations than daytime maxima. For a crop whose yield potential is set by the interaction of temperature, radiation, and water availability during a finite window, every day shaved off that window is a day of lost photosynthesis and lost grain.

The researchers are explicit about the adaptation pathway. Farmers in the affected highlands, they argue, must switch to early-maturing, drought-tolerant crop varieties and implement soil moisture conservation practices. Early-maturing wheat cultivars can complete their life cycle within the shortened and increasingly unpredictable growing window, escaping terminal drought that longer-season varieties cannot. Drought-tolerant genetics, meanwhile, maintain yield under the moisture deficits that 58 percent below-average rainfall years make routine rather than exceptional. Soil moisture conservation, including practices such as mulching, tied ridging, and improved infiltration management, addresses the other side of the water balance, capturing the intense, concentrated rainfall episodes that the Precipitation Concentration Index revealed and holding that water in the root zone for the crop to use.

What makes this study valuable is its methodological transparency and its grounding in three decades of station data rather than coarse gridded products. By combining the Mann–Kendall and Sen’s slope tests for trend, multiple variability indices for character, and regression for impact, the researchers built a chain of evidence that runs unbroken from the rain gauge to the wheat field. The picture that emerges is of a farming system under compound stress: rainfall that is late, concentrated, and frequently below normal; temperatures rising on both daily extremes; growing seasons that contract and shift; and yields that in some places swing almost entirely with the weather. For the millions of smallholders who depend on bread wheat in the Ethiopian highlands, the study is less a forecast than a description of the present, and its central message is that adaptation choices, in variety, in planting strategy, and in soil management, will determine whether the region’s most important staple can keep pace with its changing climate.

Subject of Research: Climate variability and its impact on bread wheat productivity in the central highlands of Ethiopia

Article Title: Analysis of climate trend, variability and its implications on bread wheat (Triticum aestivum L.) productivity in central highlands of Ethiopia

Article References: Hunduma, S., Meseret, A., Edao, A. L., & Yoseph, T. (2026). Analysis of climate trend, variability and its implications on bread wheat (Triticum aestivum L.) productivity in central highlands of Ethiopia. Discover Sustainability. https://doi.org/10.1007/s43621-026-04710-8

Image Credits: AI Generated

DOI: 10.1007/s43621-026-04710-8

Keywords: climate variability, bread wheat, Ethiopia, rainfall trends, temperature trends, growing season, Mann-Kendall test, food security, agrometeorology, crop adaptation, highlands, drought

The Mediterranean basin is widely regarded as one of the planet’s most vulnerable climate hotspots, a region where warming is progressing faster than the global average and where water resources are already under strain. Understanding how temperature and rainfall have behaved in the past is essential for projecting what comes next, yet the observational record has always been the weak link. Individual weather stations open and close, instruments are replaced, stations move, and urban development alters local readings. These discontinuities mean that a simple average of available measurements can mislead, and the further back in time one looks, the sparser and less reliable the network becomes. Traditional approaches to this problem rely on homogenisation techniques, statistical procedures designed to detect and correct artificial shifts in the data, but these methods struggle when station density drops or when the inhomogeneities themselves are difficult to distinguish from genuine climate signals.

The Pisa team’s answer was to let machine learning do the heavy lifting. At the core of the methodology are two families of deep learning models: Long Short-Term Memory networks, known as LSTMs, and fully connected neural networks. LSTMs are a specialised form of recurrent neural network introduced in the late 1990s, designed specifically to learn patterns in sequential data. Unlike conventional networks, they carry information forward through time using internal memory cells and gating mechanisms, which allows them to capture dependencies that stretch across months and seasons. This makes them naturally suited to climate data, where the value of a variable in one month is shaped by a chain of preceding conditions. The fully connected networks complement this by learning nonlinear relationships between predictors at each location, drawing on spatial context to estimate values where direct observations are missing.

Trained on the extensive Italian observational archive, the models learned to reconstruct monthly climate variables across the regular 10 kilometre grid, effectively weaving thousands of patchy station records into a single seamless fabric. The output is a set of climate fields that are coherent both in space and in time, meaning that neighbouring grid cells tell a physically plausible story and that the evolution of any given cell through the decades is smooth rather than erratic. This is a crucial property, because naive gap-filling can produce artefacts, sudden jumps or dips that reflect the data rescue process rather than the actual atmosphere. The reconstructed product then had to prove itself against an independent benchmark, and the researchers turned to ERA5, the state-of-the-art global reanalysis dataset produced by the European Centre for Medium-Range Weather Forecasts, which blends observations with a numerical weather model to create a physically consistent record of the atmosphere.

The validation results are striking. Correlations between the reconstructed fields and ERA5 exceed 0.96 for temperature variables, an agreement so close that the two datasets are nearly indistinguishable in their broad patterns. For cumulative precipitation, correlations exceed 0.8, which is a strong result given that precipitation is notoriously harder to model than temperature, being patchier in space, more episodic in time, and heavily influenced by local topography. In a country like Italy, where the Alps and the Apennines create dramatic orographic effects and where a few tens of kilometres can separate a humid coastal plain from a rain-shadow valley, capturing precipitation realistically is a formidable challenge. The fact that a data-driven model achieves this level of fidelity suggests that the observational archive, once properly exploited, contains enough information to constrain the climate signal even where individual stations are silent.

With a validated reconstruction in hand, the team turned to the question that motivated the work: what has actually happened to Italy’s climate since 1950? The trend analysis revealed three key indicators of ongoing change. The first is widespread and persistent warming, with rates exceeding 0.04 degrees Celsius per year in mountainous regions. That figure, equivalent to more than four degrees per century, underscores the particular exposure of high-altitude environments, where snowpack, glaciers, and seasonal water storage are all sensitive to even modest temperature shifts. Mountain areas act as early warning systems for the wider climate system, and the Italian Alps and Apennines are clearly sounding the alarm.

The second and third indicators concern rainfall, and together they tell a more subtle and arguably more consequential story. The analysis found a significant decline in monthly cumulative rainfall across the country, meaning that the total volume of water delivered over a month has been shrinking. At the same time, it documented an intensification of daily extreme rainfall events, meaning that when it does rain, the downpours are becoming heavier. This dual pattern, less widespread rainfall but more intense extremes, is precisely the signature that physical reasoning predicts for a warming world. A warmer atmosphere holds more water vapour, roughly seven percent more per degree of warming, following the Clausius-Clapeyron relation, so any given storm has more moisture to draw upon. Meanwhile, changes in synoptic-scale atmospheric circulation, the large-scale patterns that steer weather systems across the Mediterranean, can reduce the frequency of rain-bearing conditions, so that dry spells lengthen between the increasingly violent events.

The authors interpret this combination as evidence of a structural transformation of the Italian hydrological cycle, driven both by thermodynamic processes, the direct physics of a warmer, moister atmosphere, and by shifts in the circulation patterns that govern when and where rain falls. The implications are far-reaching. A hydrological cycle that delivers the same or less total water in fewer, more concentrated bursts is a cycle that produces both more droughts and more floods. Reservoirs and soils that once received steady replenishment now face long dry intervals punctuated by runoff events so intense that much of the water escapes quickly to the sea rather than recharging aquifers. Agriculture, hydropower, civil protection, and urban drainage systems were all designed around a rainfall regime that the data show is no longer stable.

What elevates this study beyond a national climate audit is the accessibility of its output. The reconstructed dataset has been made publicly available through the AIClimate platform hosted by the University of Pisa, offering a concrete resource for climate change studies, hydrological modelling, and the planning of adaptation strategies in highly vulnerable regions. For hydrologists, a coherent 70-year gridded record is invaluable for calibrating rainfall-runoff models and assessing flood and drought hazard. For planners and policymakers, it provides the historical baseline against which future projections can be judged and against which adaptation measures, from reservoir management to land-use planning, can be designed. And for climate scientists studying the Mediterranean, it removes a long-standing obstacle: the fragmentation of national archives that has hampered regional synthesis.

The study also sits within a broader movement in the geosciences, one in which deep learning is increasingly deployed not to replace physical models but to extract maximum value from imperfect observations. From filling gaps in rainfall databases to predicting floods in fast-flowing watersheds and forecasting river flow in small catchments, the same family of techniques is proving adept at problems where data exist but are messy, incomplete, or nonlinear in ways that classical statistics handle poorly. Italy, with one of the longest and richest observational histories in the world yet also one of the most heterogeneous, is an ideal proving ground. The message of this research is twofold: the past seven decades of Italian climate can now be seen whole, and what they reveal is a country where warming is entrenched in its mountains, its rains are thinning, and its storms are growing fiercer. That combination, captured with unprecedented spatial detail and validated against independent reanalysis, turns a fragmented archive into a national climate memory, and it arrives at a moment when the Mediterranean can least afford to forget.

Subject of Research: Deep learning reconstruction of monthly temperature and precipitation trends across Italy from 1950 to 2020

Article Title: Artificial intelligence for climate reconstruction: Spatiotemporal modelling of precipitation and temperature trends in Italy

Article References: Bini, M., & Luppichini, M. (2026). Artificial intelligence for climate reconstruction: Spatiotemporal modelling of precipitation and temperature trends in Italy. Theoretical and Applied Climatology, 157(10), Article 667. https://doi.org/10.1007/s00704-026-06599-9

Image Credits: AI Generated

DOI: 10.1007/s00704-026-06599-9

Keywords: artificial intelligence, deep learning, LSTM, climate reconstruction, Italy, Mediterranean climate, precipitation trends, temperature trends, ERA5 reanalysis, extreme rainfall, hydrological cycle, climate change

The research team, led by Anurag Saxena of the ICAR-National Dairy Research Institute in Karnal and including researchers based in Leh itself, relied on gridded meteorological datasets rather than a single weather station record. This choice matters in a region as topographically extreme as Ladakh, where the western Himalaya and Karakoram ranges collide and station coverage has historically been sparse. Gridded products interpolate observations across a regular spatial grid, allowing analysts to build continuous time series even where instrumentation is limited. The team examined both annual and seasonal breakdowns of temperature and precipitation, looking for departures from long-term climatological normals as well as directional trends.

To detect those trends, the researchers deployed two of the most widely used tools in hydro-climatology: the Mann-Kendall test and Sen’s slope estimator. The Mann-Kendall test, first formulated by Henry Mann in 1945 and refined by Maurice Kendall, is a non-parametric method that assesses whether a time series shows a monotonic upward or downward tendency without assuming any particular data distribution. Sen’s slope estimator complements it by providing a robust estimate of the magnitude of that trend, calculated from the median of all pairwise slopes in the data. Together, the methods were applied at the 90 percent and 95 percent confidence levels, meaning the authors only flagged a trend as statistically significant when the probability of it arising by chance fell below those thresholds.

The headline result is an asymmetric warming pattern. Rather than maximum and minimum temperatures rising in lockstep, the study found a gradual warming tendency concentrated particularly in minimum temperatures, the nighttime and winter lows. This kind of asymmetric warming is characteristic of high-altitude cold deserts and has been documented across the broader Himalayan region and the Tibetan Plateau. Elevation-dependent warming, the phenomenon whereby mountain regions warm faster than surrounding lowlands, has been attributed to a suite of feedbacks, including snow-albedo changes as reflective snow and ice give way to darker surfaces that absorb more solar radiation, shifts in cloud cover, and changes in atmospheric water vapour. Warmer nights in a cold desert may sound benign, even welcome, but they carry cascading consequences for snowpack persistence, glacier mass balance, and the timing of spring melt.

That last point is where the precipitation findings become consequential. Leh’s agriculture and its entire hydro-social system depend on meltwater from winter snow and glaciers feeding the streams that irrigate barley fields, vegetable plots, and the famous groves of apricot and willow. The study found that seasonal precipitation exhibits substantial inter-annual variability, with no statistically significant long-term trend across the full 1951 to 2024 record. In other words, some years are wet, some are parched, and the record does not yet show a consistent direction of change when the whole period is considered. However, the authors highlight a more troubling recent signal: declining winter and pre-monsoon precipitation in recent decades. Because winter snowfall and spring rains are precisely the inputs that sustain snowmelt-driven hydrology into the growing season, a decline in these seasons could tighten agricultural water availability even if annual totals appear stable.

The hydrological stakes extend far beyond Leh’s district boundaries. The Indus basin, into which Ladakh’s meltwaters drain, sustains hundreds of millions of people downstream in India and Pakistan, and research has repeatedly identified snow- and glacier-fed Asian river systems as among the most climate-vulnerable water towers on the planet. Studies of the upper Indus have long noted that its hydrological regime is dominated by cryosphere inputs rather than direct rainfall, making the timing and volume of snow accumulation in any given winter a first-order determinant of summer water supply. A warming signal that preferentially raises minimum temperatures can shorten the snow accumulation season, increase mid-winter melt events, and shift the peak of the meltwater pulse earlier in the year, desynchronising it from the irrigation calendar that Ladakhi farmers have refined over centuries.

The study’s statistical caution is itself instructive. In trend analysis of hydro-climatic data, serial correlation, the tendency of one year’s value to influence the next, can inflate the apparent significance of trends, and methods such as prewhitening or effective sample size corrections have been developed to address this. The authors’ decision to report trends at two confidence levels and to emphasise departures from normals alongside formal trend tests reflects a broader recognition that in highly variable cold-desert climates, a lack of statistical significance does not mean a lack of physical consequence. Declining winter precipitation that fails to cross a 95 percent significance threshold can still push a water system closer to its limits, particularly when it compounds with other stressors.

Those compounding stressors are well documented in the wider literature on Ladakh. The region’s glaciers have retreated measurably over the past five decades, satellite analyses have shown accelerating ice loss across the Himalayas in recent decades, and the 2010 cloudburst and flash flood that struck Leh killed hundreds and underscored the region’s exposure to hydro-meteorological extremes. At the same time, the town of Leh has urbanised rapidly, transforming traditional hydro-social relations and increasing demand on groundwater and surface irrigation networks. Pastoralist communities such as the Changpa face their own climate vulnerabilities on the high plateaus where they herd pashmina goats. Against this backdrop, a 74-year quantification of the underlying climatic trends provides an essential baseline for separating long-term directional change from the noise of natural variability.

The research was conducted under a Department of Science and Technology project aimed at developing climate-resilient and sustainable agriculture-based systems for the cold-arid region of Ladakh, and the authors’ framing reflects that applied mandate. They argue that their findings underscore the importance of climate-resilient water management, agriculture, and monitoring strategies for the district. In practice, that means investments in water storage that can buffer against erratic snowmelt, continued support for adaptive innovations such as the artificial glaciers that have been pioneered in Ladakh to store winter water as ice, and sustained meteorological monitoring to track whether the recent declines in winter and pre-monsoon precipitation harden into statistically robust trends.

For a region often described as a sentinel of global climate change, Leh’s record offers a nuanced message. The warming signal is real and detectable, and it is strongest in the minimum temperatures that govern snow survival and winter ecology. The precipitation signal, by contrast, is dominated by variability, and the most actionable recent development, a possible decline in the winter and spring moisture that feeds the meltwater economy, remains on the edge of statistical certainty. What the study makes clear is that in a cold desert where every drop of water arrives either as snow on the peaks or as a carefully timed irrigation turn, even subtle shifts in the timing and form of precipitation can matter as much as headline warming numbers. As the Himalayan cryosphere continues to change, long-term analyses of this kind will be indispensable for the farmers, herders, and town planners of Ladakh who must adapt to a climate that is warming asymmetrically and raining, and snowing, on its own unpredictable schedule.

Subject of Research: Long-term trends and variability in temperature and precipitation extremes in the high-altitude cold desert of Leh, Ladakh, India

Article Title: Trends and variability of temperature and precipitation extremes in Leh (Ladakh), India

Article References: Saxena, A., Ram, H., Meena, R. K., Pandey, S., Yadav, A., Bhattarchjee, S., Rajkumari, P., Sonkusale, L., Raghuwanshi, M. S., Stanzin, J., & Landol, S. (2026). Trends and variability of temperature and precipitation extremes in Leh (Ladakh), India. Environmental Monitoring and Assessment, 198(10), Article 1127. https://doi.org/10.1007/s10661-026-15964-0

Image Credits: AI Generated

DOI: 10.1007/s10661-026-15964-0

Keywords: climate change, Ladakh, Leh, Himalaya, temperature trends, precipitation variability, Mann-Kendall test, Sen's slope estimator, cold desert, snowmelt hydrology, elevation-dependent warming, water resources